Published: 2019-07-08 21:10:00
By ytwan
In Big Data .
tags: Flink
1.Flink Type_Information
2.Flink中数据
3.Flink的数据类型
4.类型
5.Flink Java Tuples
Tuple 的构建：
01 . Tuple -- 抽象Build类 - 构建与产品类相关的部件
public abstract class Tuple implements java . io . Serializable {
public static final int
public abstract < T > T getField ( int pos );
public < T > T getFieldNotNull ( int pos ){}
public abstract < T > void setField ( T value , int pos );
public abstract int getArity ();
public abstract < T extends Tuple > T copy ();
public static Class <? extends Tuple > getTupleClass ( int arity ) {}
public static Tuple newInstance ( int arity ) {}
private static final Class < ?> [] CLASSES = new Class <? > [] {}
}
02. Tuple2 -- 抽象Build的实现类 - 真正构建产品的类
public class Tuple2 < T0 , T1 > extends Tuple {
public T0 f0 ;
public T1 f1 ;
public Tuple2 () {}
public int getArity () {
return 2 ;
}
public < T > T getField ( int pos )
public < T > void setField ( T value , int pos )
public void setFields ( T0 value0 , T1 value1 ) {}
public String toString () {
public boolean equals ( Object o ) {
public int hashCode () {
public static < T0 , T1 > Tuple2 < T0 , T1 > of ( T0 value0 , T1 value1 ){
return new Tuple2 <> ( value0 , value1 );
}
}
03. Tuple2Builder Director -- 将builder构建的产品部件组装成产品 , 对外隐藏内部组装细节
public class Tuple2Builder < T0 , T1 > {
private List < Tuple2 < T0 , T1 >> tuples = new ArrayList <> ();
public Tuple2Builder < T0 , T1 > add ( T0 value0 , T1 value1 ){
tuples . add ( new Tuple2 <> ( value0 , value1 ));
return this ;
}
@ SuppressWarnings ( "unchecked" )
public Tuple2 < T0 , T1 > [] build (){
return tuples . toArray ( new Tuple2 [ tuples . size ()]);
}
}
04. main 使用
List < Tuple2 < String , Integer >> input1 = Arrays . asList (
new Tuple2Builder < String , Integer > ()
. add ( "foo" , 1 )
. add ( "foobar" , 1 )
. add ( "foo" , 1 )
. add ( "bar" , 1 )
. add ( "foo" , 1 )
. add ( "foo" , 1 )
. build ()
updateEmpty = mapOf ( Tuple2 . of ( 3 , "3" ), Tuple2 . of ( 5 , "5" ), Tuple2 . of ( 23 , null ), Tuple2 . of ( 10 , "10" ));
说明：
面对的场景：
各个子对象组合出这个复杂对象的过程相对来说比较稳定，但是子对象的创建过程各不相同并且可能面临变化
解决的方式
将配置从目标类中隔离出来，避免作为过多的setter方法
Builder模式比较常见的实现形式是通过链式调用
-- 复杂对象的 构建 与它的 表示 分离
TupleTypeInfo
public final class TupleTypeInfo < T extends Tuple > extends TupleTypeInfoBase < T > {
public TupleTypeInfo ( TypeInformation <? >... types ) {
this (( Class < T > ) Tuple . getTupleClass ( types . length ), types );
}
getFieldNames
getFieldIndex
createSerializer
private class TupleTypeComparatorBuilder
getGenericParameters
getBasicTupleTypeInfo
getBasicAndBasicValueTupleTypeInfo
}
public abstract class TupleTypeInfoBase < T > extends CompositeType < T >
public abstract class CompositeType < T > extends TypeInformation < T >
public abstract class TypeInformation < T > implements Serializable
TupleSerializer
public class TupleSerializer < T extends Tuple > extends TupleSerializerBase < T > implements SelfResolvingTypeSerializer < T > {
TupleSerializerBase
public abstract class TupleSerializerBase < T > extends TypeSerializer < T > {
Tuple：
Flink tuples 是固定长度固定类型的Java Tuple实现
TupleTypeInfo 任意的Flink tuple 类型 ( 支持Tuple1 to Tuple25 )
package org . apache . flink . api . java . tuple ;
package org . apache . flink . api . java . typeutils ;
package org . apache . flink . api . common . typeutils ;
package org . apache . flink . api . common . typeinfo ;
Java 和Scala
CaseClassTypeInfo : 任意的 Scala CaseClass ( 包括 Scala tuples ) 。
Flink的TypeInformation类
TypeInformation类是所有类型描述类的基类，它揭示了数据类型的一些基础属性，并产生序列化器在默写特殊情况下生成类型比较器
相关语法
参考：